| 1 | package cdp |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "errors" |
| 7 | "fmt" |
| 8 | "strconv" |
| 9 | "strings" |
| 10 | "sync" |
| 11 | |
| 12 | "reasonix/internal/browser" |
| 13 | ) |
| 14 | |
| 15 | // worldName labels the isolated world in DevTools; page script can neither |
| 16 | // enumerate nor reach it. |
| 17 | const worldName = "reasonix-browser" |
| 18 | |
| 19 | // snapshotBudget caps the nodes one snapshot may emit so a large document |
| 20 | // cannot flood the model's context. |
| 21 | const snapshotBudget = 2000 |
| 22 | |
| 23 | // pageIdentity is fixed for a tab's whole life: which target it is, which |
| 24 | // session may drive it, and which partition it belongs to. |
| 25 | type pageIdentity struct { |
| 26 | id string |
| 27 | target string |
| 28 | session string |
| 29 | context string |
| 30 | owner string |
| 31 | temporary bool |
| 32 | } |
| 33 | |
| 34 | // pageDocument is what one document version owns. Navigation, page |
| 35 | // replacement, and a take-over retire all of it together, which is why it is |
| 36 | // one value and not five independent flags: no combination of a live token |
| 37 | // with a dead isolated world can exist. |
| 38 | type pageDocument struct { |
| 39 | frame string |
| 40 | world int64 |
| 41 | token string |
| 42 | lastSeq int64 |
| 43 | takenOver bool |
| 44 | } |
| 45 | |
| 46 | // page is one agent-owned tab. Its mutex guards the document state and the |
| 47 | // per-tab records that outlive a single document. |
| 48 | type page struct { |
| 49 | pageIdentity |
| 50 | detach []func() |
| 51 | |
| 52 | mu sync.Mutex |
| 53 | doc pageDocument |
| 54 | loading bool |
| 55 | dead bool |
| 56 | url string |
| 57 | title string |
| 58 | downloads []string |
| 59 | } |
| 60 | |
| 61 | // retire drops everything bound to the document that just went away: the |
| 62 | // isolated world holding the refs, the token those refs were promised under, |
| 63 | // and the take-over counter the new world starts over from. The frame outlives |
| 64 | // the document that was loaded in it. |
| 65 | func (p *page) retire(url string) { |
| 66 | p.mu.Lock() |
| 67 | defer p.mu.Unlock() |
| 68 | p.doc = pageDocument{frame: p.doc.frame} |
| 69 | if url != "" { |
| 70 | p.url = url |
| 71 | } |
| 72 | } |
| 73 | |
| 74 | func (p *page) tab() browser.Tab { |
| 75 | p.mu.Lock() |
| 76 | defer p.mu.Unlock() |
| 77 | return browser.Tab{ID: p.id, URL: p.url, Title: p.title, Loading: p.loading, Temporary: p.temporary} |
| 78 | } |
| 79 | |
| 80 | func (p *page) setLoading(loading bool) { |
| 81 | p.mu.Lock() |
| 82 | p.loading = loading |
| 83 | p.mu.Unlock() |
| 84 | } |
| 85 | |
| 86 | func (p *page) markDead() { |
| 87 | p.mu.Lock() |
| 88 | p.dead = true |
| 89 | p.doc.world = 0 |
| 90 | p.doc.token = "" |
| 91 | p.mu.Unlock() |
| 92 | } |
| 93 | |
| 94 | func (p *page) isDead() bool { |
| 95 | p.mu.Lock() |
| 96 | defer p.mu.Unlock() |
| 97 | return p.dead |
| 98 | } |
| 99 | |
| 100 | // attach opens a flat DevTools session on target and wires the events that |
| 101 | // retire document state. The returned page is not yet bootstrapped: the |
| 102 | // isolated world is created lazily by the first snapshot. |
| 103 | func (e *Executor) attach(ctx context.Context, id, target, contextID, owner string, temporary bool) (*page, error) { |
| 104 | var out struct { |
| 105 | SessionID string `json:"sessionId"` |
| 106 | } |
| 107 | if err := e.conn.call(ctx, "", "Target.attachToTarget", map[string]any{"targetId": target, "flatten": true}, &out); err != nil { |
| 108 | return nil, fmt.Errorf("attach to tab: %w", err) |
| 109 | } |
| 110 | p := &page{pageIdentity: pageIdentity{ |
| 111 | id: id, target: target, session: out.SessionID, context: contextID, owner: owner, temporary: temporary, |
| 112 | }} |
| 113 | if err := e.conn.call(ctx, p.session, "Page.enable", nil, nil); err != nil { |
| 114 | return nil, fmt.Errorf("enable page events: %w", err) |
| 115 | } |
| 116 | e.subscribe(p) |
| 117 | e.refreshFrame(ctx, p) |
| 118 | return p, nil |
| 119 | } |
| 120 | |
| 121 | // subscribe binds the page's document lifetime to Chrome's frame events. |
| 122 | func (e *Executor) subscribe(p *page) { |
| 123 | p.detach = append(p.detach, |
| 124 | e.conn.on(p.session, "Page.frameNavigated", func(params json.RawMessage) { |
| 125 | var ev struct { |
| 126 | Frame struct { |
| 127 | ID string `json:"id"` |
| 128 | ParentID string `json:"parentId"` |
| 129 | URL string `json:"url"` |
| 130 | } `json:"frame"` |
| 131 | } |
| 132 | if err := json.Unmarshal(params, &ev); err != nil || ev.Frame.ParentID != "" { |
| 133 | return |
| 134 | } |
| 135 | p.mu.Lock() |
| 136 | p.doc.frame = ev.Frame.ID |
| 137 | p.mu.Unlock() |
| 138 | p.retire(ev.Frame.URL) |
| 139 | }), |
| 140 | e.conn.on(p.session, "Page.frameStartedLoading", func(json.RawMessage) { p.setLoading(true) }), |
| 141 | e.conn.on(p.session, "Page.frameStoppedLoading", func(json.RawMessage) { p.setLoading(false) }), |
| 142 | e.conn.on(p.session, "Runtime.executionContextsCleared", func(json.RawMessage) { p.retire("") }), |
| 143 | e.conn.on(p.session, "Inspector.targetCrashed", func(json.RawMessage) { p.markDead() }), |
| 144 | ) |
| 145 | } |
| 146 | |
| 147 | func (p *page) unsubscribe() { |
| 148 | for _, off := range p.detach { |
| 149 | off() |
| 150 | } |
| 151 | p.detach = nil |
| 152 | } |
| 153 | |
| 154 | // refreshFrame records the main frame and the tab's current address. |
| 155 | func (e *Executor) refreshFrame(ctx context.Context, p *page) { |
| 156 | var tree struct { |
| 157 | FrameTree struct { |
| 158 | Frame struct { |
| 159 | ID string `json:"id"` |
| 160 | URL string `json:"url"` |
| 161 | Title string `json:"title"` |
| 162 | } `json:"frame"` |
| 163 | } `json:"frameTree"` |
| 164 | } |
| 165 | if err := e.conn.call(ctx, p.session, "Page.getFrameTree", nil, &tree); err != nil { |
| 166 | return |
| 167 | } |
| 168 | p.mu.Lock() |
| 169 | p.doc.frame = tree.FrameTree.Frame.ID |
| 170 | if p.url == "" { |
| 171 | p.url = tree.FrameTree.Frame.URL |
| 172 | } |
| 173 | p.mu.Unlock() |
| 174 | } |
| 175 | |
| 176 | // evalResult is the shape of a Runtime.evaluate or Runtime.callFunctionOn |
| 177 | // reply that this package cares about. |
| 178 | type evalResult struct { |
| 179 | Result struct { |
| 180 | Type string `json:"type"` |
| 181 | Subtype string `json:"subtype"` |
| 182 | Value json.RawMessage `json:"value"` |
| 183 | ObjectID string `json:"objectId"` |
| 184 | } `json:"result"` |
| 185 | ExceptionDetails *struct { |
| 186 | Text string `json:"text"` |
| 187 | Exception *struct { |
| 188 | Description string `json:"description"` |
| 189 | } `json:"exception"` |
| 190 | } `json:"exceptionDetails"` |
| 191 | } |
| 192 | |
| 193 | func (r evalResult) exception() error { |
| 194 | if r.ExceptionDetails == nil { |
| 195 | return nil |
| 196 | } |
| 197 | if r.ExceptionDetails.Exception != nil && r.ExceptionDetails.Exception.Description != "" { |
| 198 | return fmt.Errorf("page script failed: %s", firstLine(r.ExceptionDetails.Exception.Description)) |
| 199 | } |
| 200 | return fmt.Errorf("page script failed: %s", r.ExceptionDetails.Text) |
| 201 | } |
| 202 | |
| 203 | func firstLine(s string) string { |
| 204 | line, _, _ := strings.Cut(s, "\n") |
| 205 | return line |
| 206 | } |
| 207 | |
| 208 | // ensureWorld returns the page's isolated world, creating and bootstrapping it |
| 209 | // when the previous document took the old one with it. |
| 210 | func (e *Executor) ensureWorld(ctx context.Context, p *page) (int64, error) { |
| 211 | p.mu.Lock() |
| 212 | world, frame := p.doc.world, p.doc.frame |
| 213 | p.mu.Unlock() |
| 214 | if world != 0 { |
| 215 | return world, nil |
| 216 | } |
| 217 | if frame == "" { |
| 218 | e.refreshFrame(ctx, p) |
| 219 | p.mu.Lock() |
| 220 | frame = p.doc.frame |
| 221 | p.mu.Unlock() |
| 222 | } |
| 223 | if frame == "" { |
| 224 | return 0, fmt.Errorf("tab %s has no main frame", p.id) |
| 225 | } |
| 226 | var out struct { |
| 227 | ExecutionContextID int64 `json:"executionContextId"` |
| 228 | } |
| 229 | if err := e.conn.call(ctx, p.session, "Page.createIsolatedWorld", map[string]any{"frameId": frame, "worldName": worldName}, &out); err != nil { |
| 230 | return 0, fmt.Errorf("create isolated world: %w", err) |
| 231 | } |
| 232 | var boot evalResult |
| 233 | if err := e.conn.call(ctx, p.session, "Runtime.evaluate", evaluateParams(bootstrapJS, out.ExecutionContextID, true), &boot); err != nil { |
| 234 | return 0, fmt.Errorf("install page helper: %w", err) |
| 235 | } |
| 236 | if err := boot.exception(); err != nil { |
| 237 | return 0, err |
| 238 | } |
| 239 | p.mu.Lock() |
| 240 | p.doc.world = out.ExecutionContextID |
| 241 | p.doc.lastSeq = 0 |
| 242 | p.mu.Unlock() |
| 243 | return out.ExecutionContextID, nil |
| 244 | } |
| 245 | |
| 246 | func evaluateParams(expression string, contextID int64, byValue bool) map[string]any { |
| 247 | return map[string]any{ |
| 248 | "expression": expression, |
| 249 | "contextId": contextID, |
| 250 | "returnByValue": byValue, |
| 251 | "awaitPromise": true, |
| 252 | "userGesture": true, |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | // eval runs one expression in the page's isolated world and decodes its value. |
| 257 | // A world that died between the lookup and the call is rebuilt once, which is |
| 258 | // the ordinary race with a page navigating itself. |
| 259 | func (e *Executor) eval(ctx context.Context, p *page, expression string, out any) error { |
| 260 | for attempt := range 2 { |
| 261 | world, err := e.ensureWorld(ctx, p) |
| 262 | if err != nil { |
| 263 | return err |
| 264 | } |
| 265 | var res evalResult |
| 266 | err = e.conn.call(ctx, p.session, "Runtime.evaluate", evaluateParams(expression, world, out != nil), &res) |
| 267 | if err != nil { |
| 268 | if attempt == 0 && staleContext(err) { |
| 269 | p.retire("") |
| 270 | continue |
| 271 | } |
| 272 | return err |
| 273 | } |
| 274 | if err := res.exception(); err != nil { |
| 275 | return err |
| 276 | } |
| 277 | if out == nil { |
| 278 | return nil |
| 279 | } |
| 280 | if len(res.Result.Value) == 0 { |
| 281 | return fmt.Errorf("page helper returned no value") |
| 282 | } |
| 283 | return json.Unmarshal(res.Result.Value, out) |
| 284 | } |
| 285 | return fmt.Errorf("tab %s replaced its document while the call was in flight", p.id) |
| 286 | } |
| 287 | |
| 288 | // evalHandle runs one expression and keeps the remote object alive so a DOM |
| 289 | // command can address the node it returned. |
| 290 | func (e *Executor) evalHandle(ctx context.Context, p *page, expression string) (string, error) { |
| 291 | world, err := e.ensureWorld(ctx, p) |
| 292 | if err != nil { |
| 293 | return "", err |
| 294 | } |
| 295 | var res evalResult |
| 296 | if err := e.conn.call(ctx, p.session, "Runtime.evaluate", evaluateParams(expression, world, false), &res); err != nil { |
| 297 | return "", err |
| 298 | } |
| 299 | if err := res.exception(); err != nil { |
| 300 | return "", err |
| 301 | } |
| 302 | if res.Result.Subtype == "null" || res.Result.ObjectID == "" { |
| 303 | return "", browser.ErrStaleReference |
| 304 | } |
| 305 | return res.Result.ObjectID, nil |
| 306 | } |
| 307 | |
| 308 | func (e *Executor) releaseHandle(ctx context.Context, p *page, objectID string) { |
| 309 | if objectID == "" { |
| 310 | return |
| 311 | } |
| 312 | _ = e.conn.call(ctx, p.session, "Runtime.releaseObject", map[string]any{"objectId": objectID}, nil) |
| 313 | } |
| 314 | |
| 315 | func staleContext(err error) bool { |
| 316 | var pe *protocolError |
| 317 | if !errors.As(err, &pe) { |
| 318 | return false |
| 319 | } |
| 320 | msg := strings.ToLower(pe.Message) |
| 321 | return strings.Contains(msg, "cannot find context") || strings.Contains(msg, "execution context was destroyed") |
| 322 | } |
| 323 | |
| 324 | // pageState is the isolated world's view of the document between calls. |
| 325 | type pageState struct { |
| 326 | UserSeq int64 `json:"userSeq"` |
| 327 | URL string `json:"url"` |
| 328 | Title string `json:"title"` |
| 329 | Ready string `json:"ready"` |
| 330 | } |
| 331 | |
| 332 | // observe reads the document's state and decides whether a human touched the |
| 333 | // page since the last snapshot. The take-over flag is sticky: only a fresh |
| 334 | // snapshot clears it, matching the shell's rule that the agent resumes after |
| 335 | // re-reading the page. |
| 336 | func (e *Executor) observe(ctx context.Context, p *page) (pageState, error) { |
| 337 | var st pageState |
| 338 | if err := e.eval(ctx, p, "__rx.state()", &st); err != nil { |
| 339 | return pageState{}, err |
| 340 | } |
| 341 | p.mu.Lock() |
| 342 | defer p.mu.Unlock() |
| 343 | p.url, p.title = st.URL, st.Title |
| 344 | if st.UserSeq != p.doc.lastSeq { |
| 345 | p.doc.lastSeq = st.UserSeq |
| 346 | p.doc.takenOver = true |
| 347 | } |
| 348 | return st, nil |
| 349 | } |
| 350 | |
| 351 | // markAgentInput opens the window in which trusted input is the executor's own |
| 352 | // rather than the user's. |
| 353 | func (e *Executor) markAgentInput(ctx context.Context, p *page, windowMillis int) error { |
| 354 | var seq int64 |
| 355 | if err := e.eval(ctx, p, "__rx.window("+strconv.Itoa(windowMillis)+")", &seq); err != nil { |
| 356 | return err |
| 357 | } |
| 358 | p.mu.Lock() |
| 359 | p.doc.lastSeq = seq |
| 360 | p.mu.Unlock() |
| 361 | return nil |
| 362 | } |
| 363 |